Google Is Quietly Inventing a New Media Metric: Contribution

Google is experimenting with a strange new idea: paying publishers not for traffic, but for how much their content helps produce an AI answer.

According to Digiday, its “AI contribution” pilot compensates participating publishers when their content significantly contributes to responses across products including Gemini, AI Overviews and AI Mode. Publishers can reportedly see earnings figures in Search Console, although Google has not made the underlying calculation transparent.

At first glance, this looks like another publisher compensation programme. The more interesting possibility is that Google is testing whether informational contribution itself can become a measurable economic unit.

That is a much harder problem than paying for a click or licensing a corpus. An AI answer might draw on five publishers, use one decisive fact from each, cite some sources but not others, and still produce a recommendation or decision worth real money.

The difficult question is what, exactly, counts as contribution.

Contribution is not the same as citation

The simplest approach would be to reward whatever gets cited, but citation is a poor proxy for value.

A source can be cited prominently while contributing little beyond background context. Another might supply the crucial fact that changes the answer without receiving the most visible attribution. Originality complicates things further: one publication may spend weeks producing an investigation, while several others quickly rewrite its findings and make them easier for an AI system to retrieve.

This is why contribution is a more difficult concept than citation or retrieval.

Emerging infrastructure such as SPUR’s Content Telemetry Standard begins to distinguish between stages such as retrieval, grounding, citation and presentation. That distinction matters because content being fetched is not the same as content shaping the answer.

A page can be retrieved without affecting the conclusion. A source can influence an answer without being prominently cited. And a citation alone tells us very little about how much the source mattered.

Who gets credit when five sources make one answer?

Consider someone asking an AI assistant which mortgage is best for a first-time buyer.

One financial publication has compared current rates across major lenders. Another has investigated hidden fees. A consumer finance site has analysed approval criteria, while a newspaper has reported on recent changes to lending rules. A fifth publisher has summarised much of that information into a simple guide.

The AI may use all five before recommending two options, but those sources have not contributed equally.

This starts to look less like traditional web analytics and more like a royalty problem. Several sources contribute to one output, and some mechanism has to decide how much weight each one deserves.

Should the publication that uncovered the hidden fees receive more than the site that summarised them? Should one decisive piece of reporting be worth more than general context? Should specialist expertise carry a premium?

Once platforms start paying for contribution, these stop being abstract questions. Someone has to decide what information is worth.

The real issue is who sets the price

There is another problem with contribution-based economics: measurement power.

If an AI platform retrieves publisher content, decides how much that content influenced the answer, calculates the resulting value and then pays the publisher, the entire market is effectively being priced by the buyer.

Publishers may see an earnings figure without knowing why one article was considered more valuable than another.

That is a fragile basis for a market.

A contribution model only becomes credible if publishers can understand and eventually verify how value is being calculated. Otherwise, “contribution” risks becoming another opaque platform metric: commercially important, but impossible to audit.

This may prove more consequential than the payment levels in Google’s pilot itself. The long-term question is not simply whether publishers get paid, but whether platforms get to define contribution unilaterally.

Access and contribution are different markets

Much of the publisher-AI debate so far has focused on access: whether models can crawl, train on or retrieve content, whether a licence exists, and what that access should cost.

Those are important questions, but they describe only the entrance to the system.

Contribution asks what happens afterwards.

Two articles might both be licensed and retrieved. One may have almost no effect on the resulting answer, while the other repeatedly provides the evidence an AI uses to make recommendations.

Charging the same way for both assumes access is where the value is created.

That may not always be true.

For publishers, this could eventually produce two distinct commercial layers: one around the right to access information, and another around the value generated when that information materially contributes to an output.

A specialist review or original investigation might be used relatively infrequently but carry enormous weight when it is used. A flat access fee cannot capture that very well.

Contribution potentially can.

Why advertisers should care

This is not only a publisher licensing problem.

If AI systems increasingly help people compare products and services, then publisher information becomes part of the evidence determining which brands are considered, recommended or excluded.

Imagine a student asking an AI assistant to choose between a MacBook Air, Dell XPS 13 and Lenovo Yoga for university. A tech publication finds that the MacBook Air has the strongest battery life, another review says the Dell performs better for heavier multitasking, while a consumer site flags recurring repair complaints around a particular Lenovo configuration. The AI uses those sources to narrow the options and recommends the MacBook Air and Dell.

For Apple, Dell and Lenovo, the important question is not simply whether their brands appeared in the answer. It is which pieces of information caused one product to be recommended and another to be dropped from the shortlist.

A brand could appear frequently in AI responses but still lose at the point of consideration if the evidence around it is weak. Another might be mentioned less often overall but consistently benefit from authoritative reviews at high-intent moments.

Those outcomes are commercially very different.

If contribution becomes measurable for publishers, advertisers may eventually ask the same question from the other side: which sources contributed most to our inclusion in the decision?

Contribution could change what publishers optimise for

There is a risk here too.

Once contribution becomes measurable and monetisable, publishers will naturally optimise for it.

That could reward high-quality specialist reporting, original research and authoritative reviews. But it could also create a new version of the SEO incentive problem, with publishers restructuring content around whatever makes it easiest for AI systems to retrieve and reuse.

Only this time, publishers would not be optimising for ranking.

They would be optimising for influence on the answer itself.

That makes contribution metrics more than a measurement problem. Metrics shape behaviour.

If platforms reward originality and authority, they could strengthen the economics of quality publishing. If they reward whatever is easiest to extract and reproduce, they could accelerate commoditisation.

A new unit of media value

Google’s programme remains experimental, and it is too early to know whether it will scale or whether contribution-based compensation will become common across AI platforms.

But the idea behind it is important.

Publishing has long had familiar units of value: subscriptions, impressions, clicks, affiliate conversions and licensing fees. AI introduces something less visible.

A piece of information can shape an answer, recommendation or decision without ever becoming the destination.

If that influence can be measured, it can be priced.

And if it can be priced, contribution may become a new layer of the media economy.

The important question may no longer be whether AI used a publisher’s content.

It may be how much that content changed the answer.